In this study, knowledge and performance of CHWs on malaria did not differ significantly between the dual- and single-illness management arms. Both arms had fairly high scores on knowledge of malaria (72% dual-, 70% single-); and high scores on prescribing by case scenarios (80% in both arms) and record reviews (99% dual-, 100% single-). The care received from CHWs was rated highly by caregivers in both arms (90% dual-, 84% single-). The factors perceived to influence CHW performance included community, CHW programme and health facility-related factors.
The similarity of the knowledge on malaria in the dual- and single-illness management arms suggests that the requirement to have knowledge of pneumonia may not impact negatively on knowledge of malaria. This could be due to long familiarity with malaria for this community, an argument supported by the FGD findings where the CHWs felt that training for malaria was adequate because malaria was not new in their community. The management of malaria at the community level has been in effect in Uganda since 2002 under the home-based management of fever strategy
Although most CHWs mentioned fever as a sign of malaria (93% dual-, 100% single-), knowledge of other signs was quite low. Most CHWs mentioned ITNs among the malaria prevention methods but other methods were not well known. Knowledge of malaria transmission and danger signs was high. The results are comparable to a study among village malaria workers in Cambodia where knowledge of malaria signs was low
. However, in contrast to that study where only 19.5% knew malaria transmission, more than 90% of CHWs in the current study knew how malaria is transmitted. This considerable knowledge of CHWs on malaria prevention and transmission can be utilized in strengthening malaria prevention dissemination.
The performance of CHWs in the management of malaria from case scenarios was in agreement with that from knowledge assessment, showing no difference between the dual- and single-illness management arms (median 65% for both arms). This similarly implies that pneumonia management may not impact malaria management negatively. The performance on eliciting signs and symptoms for malaria based on case scenarios was generally low (50% for both arms) and that for combined malaria and pneumonia in the dual-illness arm was lower (25%). The findings in the current study are similar to those reported in a study in Kenya where the performance of CHWs in eliciting signs for malaria over three evaluations was 41-64%
. The findings of lower performance in eliciting signs and symptoms among children with pneumonia are similar to what was found in Kenya where the sensitivity of CHW classification of pneumonia was 31.5 to 54.5%
As reported previously, the assessment of respiratory symptoms was difficult for the CHWs. This is supported by the reports from the FGDs that difficulties were experienced during the training for respiratory assessments and the quantitative findings where a higher proportion (although non-significant) of CHWs in the dual- management arm stated that the training received was not adequate. In addition, CHWs have been used to manage fever as a sign of malaria in Uganda since 2002
 and they are, therefore, more familiar with its signs and symptoms. In contrast, awareness of pneumonia is fairly new among the general population and its diagnosis involves a more complicated algorithm of counting breathing rates, assessment for noisy breathing and chest in-drawings. These tasks have been shown to create challenges
[11, 12]. The fewer number of cases of pneumonia treated by CHWs compared to malaria cases also provides fewer opportunities to improve skills in the treatment of pneumonia.
The high scores on correct prescriptions for malaria and pneumonia from case scenarios and record reviews for both the dual- and single-illness management areas are comparable to studies in Rwanda and Kenya. In Rwanda, the range of correct prescriptions for malaria was 78-99% and that for pneumonia from three districts was 85-100%
 while in Kenya the correct treatment for malaria was 91%
. In the review of records, the correct dosing of amoxicillin, though very high (96%), was significantly lower than that of anti-malarials (99%, p = 0.009) in the dual- management arm. This is probably because the amoxicillin tablets used for pneumonia treatment have not been previously used in this setting while artemether-lumefantrine (anti-malarial) has been in use since 2005
. Both artemether-lumefantrine and amoxicillin had errors in dosing mainly close to the thresholds for changes in the dose highlighting the need to emphasize the cut-off ages for the different doses during training and re-training. In addition, some children aged 12–35 months who should have received the twelve-tablet pack of amoxicillin received the six-tablet pack. This could have been due to occasional confusion of amoxicillin dosing with artemether-lumefantrine dosing since the latter has two pre-packed doses (<36 months and 36–59 months) while the former has three (<12 months, 12–35 months, and 36–59 months).
A high proportion of children recorded with fast breathing (82%) received antibiotics appropriately. This proportion is higher than what was found in another study in Uganda where only 40% of the children that needed antibiotics received them. However, similar to that study where 10% of the children without malaria or pneumonia received either antibiotics or anti-malarials, 12% of children without fast breathing in the current study received amoxicillin inappropriately
. The challenges in correct assignment of treatment to children are most likely due to difficulties in both counting and categorizing breathing rates. From observation of CHWs’ assessment of respiratory symptoms, 49% of the CHWs estimated breathing rates within five breaths of those of the doctors (gold standard) and an even smaller proportion (39%) estimated breathing rates within three units of those of the doctor. However, a higher proportion of CHWs (89%) were able to correctly categorize the breathing rates they obtained showing that the main problem may be in measurement of breathing rates. As a result, some CHWs may record inaccurate breathing rates for the sake of filling the register but instead use other criteria to treat the children. The challenges in respiratory assessment may result in misclassification of children’s illness. From observation of respiratory assessment, 14% of CHWs misclassified children as having pneumonia while 4% misclassified children as not having pneumonia compared to the gold standard assessment. This implies that although a child presenting with pneumonia symptoms will be more likely to get appropriate treatment, some children may miss treatment or be treated inappropriately. There are higher chances of children without pneumonia being treated for pneumonia than children with pneumonia not being treated for pneumonia. Children may be inappropriately treated with antibiotics due to pressure to treat with a particular drug. Nevertheless, a considerable number of children that would not have received prompt treatment for pneumonia symptoms if treatment were not integrated receive it. The proportion of CHWs with correct breathing counts in previous studies ranges from 42 to 80%
[6, 12, 27]. The findings suggest a need for better procedures and tools to assess breathing rate and provide drugs like paracetamol to the CHWs which may help them deal with the pressure to treat children that may be febrile but who do not qualify for antibiotic treatment. Additional research is also needed to determine the extent to which integrated management of childhood illness improves coverage of correct management for pneumonia.
The factors perceived to influence performance by CHW are similar to what has been reported elsewhere through mainly quantitative but also qualitative studies. Training of health workers influences their performance
. Irregular supply of drugs was found to contribute to low performance in Zambia
. Community and financial support have also been cited as influencing performance
. Community support in the form of feedback and rewards was found to have greater influence on CHW performance than that from the health system
. Large population coverage has been found to lower performance of CHWs
. Many of the factors perceived by CHWs to influence performance in the current study can be addressed through increased sensitization of the community and health workers, and improvements in the CHW programmes especially regarding the drugs and supplies management. There was community sensitization about the programme before its implementation but not afterwards thereby missing the opportunity to re-enforce messages to the community. In addition, there were monthly meetings between health workers at the health facilities in Iganga-Mayuge HDSS and the project managers of the cluster randomized trial which provided opportunity for continued sensitization of the health workers. However since the CHWs would not attend the meetings, their concerns may not have been relayed to the health workers.
The drug supplies and management could be improved at the programme level through better quantification of drug needs and timely ordering of drugs taking into account the long and varied lengths of time needed to obtain drugs from different suppliers. The long lead times experienced with some suppliers would sometimes result in drug stock outs. In addition, CHWs should be trained on identifying minimum stock levels so that they can have timely ordering of drugs. Furthermore, flexible drug replenishment systems that take into account the variations in CHWs’ patient load should be devised.
This study is limited in using knowledge tests, case scenarios and record reviews of the CHWs’ registers to assess performance because these may not reflect their actual practice. However, these methods were able to standardize the cases, present various scenarios that may not have been encountered during the study period and assess competence in applying knowledge
. These methods do not assess skills well. However, the CHWs’ skills as they assessed children for respiratory symptoms were observed. In addition, the scenarios used may not have been adequate to comprehensively assess performance in assessing signs and symptoms because the length of the questionnaire had to be limited. The combination of several methods however, offers strength to this study. About 5% of the CHWs could not be contacted. These may have had different performance from those studied, but represent only a small part of the study population.